vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-10 09:04:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

support for voice cloning is done (+2 squashed commit)

Browse source 
Squashed commit:

[e7301628] support for voice cloning is done

[1653c576] wip adding voice cloning
This commit is contained in:
Concedo 2025-03-20 23:14:10 +08:00
parent a66d0f7743
commit c1e58419c7
27 changed files with 7031 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

34
koboldcpp.py
View file

@ -310,7 +310,9 @@ class tts_load_model_inputs(ctypes.Structure):
class tts_generation_inputs(ctypes.Structure):
_fields_ = [("prompt", ctypes.c_char_p),
("speaker_seed", ctypes.c_int),
("audio_seed", ctypes.c_int)]
("audio_seed", ctypes.c_int),
("custom_speaker_text", ctypes.c_char_p),
("custom_speaker_data", ctypes.c_char_p)]
class tts_generation_outputs(ctypes.Structure):
_fields_ = [("status", ctypes.c_int),
@ -1501,11 +1503,34 @@ def tts_load_model(ttc_model_filename,cts_model_filename):
ret = handle.tts_load_model(inputs)
return ret
def tts_prepare_voice_json(jsonstr):
try:
if not jsonstr:
return None
parsed_json = json.loads(jsonstr)
txt = parsed_json.get("text","")
items = parsed_json.get("words",[])
processed = ""
if txt=="" or not items or len(items)<1:
return None
for item in items:
word = item.get("word","")
duration = item.get("duration","")
codes = item.get("codes",[])
codestr = ""
for c in codes:
codestr += f"<|{c}|>"
processed += f"{word}<|t_{duration:.2f}|><|code_start|>{codestr}<|code_end|>\n"
return {"phrase":txt.strip()+".","voice":processed.strip()}
except Exception:
return None
def tts_generate(genparams):
global args
prompt = genparams.get("input", genparams.get("text", ""))
prompt = prompt.strip()
voice = 1
speaker_json = tts_prepare_voice_json(genparams.get("speaker_json","")) #handle custom cloned voices
voicestr = genparams.get("voice", genparams.get("speaker_wav", ""))
voice_mapping = ["kobo","cheery","sleepy","shouty","chatty"]
normalized_voice = voicestr.strip().lower() if voicestr else ""
@ -1522,6 +1547,13 @@ def tts_generate(genparams):
except Exception:
aseed = -1
inputs.audio_seed = aseed
if speaker_json:
inputs.custom_speaker_text = speaker_json.get("phrase","").encode("UTF-8")
inputs.custom_speaker_data = speaker_json.get("voice","").encode("UTF-8")
inputs.speaker_seed = 100
else:
inputs.custom_speaker_text = "".encode("UTF-8")
inputs.custom_speaker_data = "".encode("UTF-8")
ret = handle.tts_generate(inputs)
outstr = ""
if ret.status==1: